Doubt about first backup speed and disabled deduplication?

[frankz]

Hi everyone, I'm trying the new proxmox backup. I added the proxmox backup server from the cluster and everything is ok. H-o performed some test backups, and I noticed that the first backup is very slow. I've done the same VMs multiple backups and the transfer snapshots work. At the moment I don't see the deduplication percentage on the datastore. I checked via zfs for a deduplication status query and resolved it off. I haven't changed anything and the PBS installation is the default one.
My questions are:
1 it is normal that from shell running the command:
zfs get all | grep dedu
ZFS_STORAGE dedup off default
ZFS_USB dedup off default

2 The system state displays significant CPU expenditure of 65> 100% and memory.

 

[oversite]

the PBS dedup has nothing to do with zfs dedup. Look at the datastore summary.

 

[frankz]

the PBS dedup has nothing to do with zfs dedup. Look at the datastore summary.

View attachment 23235

I imagined, but I still don't see anything changed, the deduplication remains at 1%. I still have to finish making the first full cluster backup so I assume deduplication kicks in after a few snapshots. Or am I wrong?

 

[frankz]

the PBS dedup has nothing to do with zfs dedup. Look at the datastore summary.

View attachment 23235

 

[oversite]

I imagined, but I still don't see anything changed, the deduplication remains at 1%. I still have to finish making the first full cluster backup so I assume deduplication kicks in after a few snapshots. Or am I wrong?

hm i dunno why, the deduplication works instantly but perhaps the status shown comes from the GC runs, so give it a few days. I am however sure that the space utilized will be deduplicated even if you don't see it yet. I am sorry but i don't know when the dedup stats is calculated.

 

[dcsapak]

but perhaps the status shown comes from the GC runs,

yes, calculating the deduplcation factor is rather cost intensive (we have to iterate over all chunks and indexes) so the only time we have that data is after a garbage collection

 

[frankz]

yes, calculating the deduplcation factor is rather cost intensive (we have to iterate over all chunks and indexes) so the only time we have that data is after a garbage collection

Hi and thanks, I was just about to write to you that I hadn't launched the collection yet so I couldn't get any deduplication results. Sorry, but being the first time I use pbs I was a bit rushed to ask for support. Thanks anyway for your valuable information.